OpenAI’s Sarah Friar on the investment needed to advance AI

As artificial intelligence models grow more advanced, the industry is “spending like crazy” on building out the underlying infrastructure, said Sarah Friar, the chief financial officer of OpenAI. “We all need more compute,” she said, referring to the chips, data centers, and other equipment that drive AI models.

Speaking at Goldman Sachs’ Disruptive Technology Symposium in London, Friar outlined the progress she envisions for AI models en route to artificial general intelligence (AGI), as well as the kinds of investment and infrastructure required to give the US an edge in the AI race.

Could an AI eventually find the proverbial cure for cancer? Friar thinks it’s a real possibility. “Professors and academics are finding that models are coming out with novel things in their field,” she said. “We don’t yet know if those novel things are real, because they now need to go and test [them]. But…we’re hearing that we might be already pushing that boundary and finding new discoveries.”

When will artificial general intelligence emerge?

Friar defined AGI as “the point at which AI systems can take on the majority of the value-added human work in the world. And we’re getting close to that being the case.”

The path to AGI involves progress by phases, Friar said. First came the AI chatbots that debuted in 2023, and that offered real-time responses to user queries. “Last year, we brought reasoning to the table,” Friar said.

The third phase, coming into view this year, involves “AI agents that can go out and do work independently for you,” Friar said. She cited models that do deep research work, perform tasks such as booking flights and holidays, or build software as examples of such AI agents that already exist. “The agentic software engineer will not just augment current software engineers in the workforce,” Friar said. “It can literally build an app for you — and not only build it, but also do its own quality assurance, bug testing, and documentation.” The next step, Friar said, would involve models that no longer just depend on human knowledge that exists in the world today; instead, they would extend that knowledge and innovate.

The industry is also spending “a lot more time on what Silicon Valley loves to call ‘vibes,’ but which is effectively EQ [emotional quotient],” Friar said. A model with the equivalent of EQ would feel more human when we interact with it, she said. Companies are also emphasizing models equipped with creative skills like writing, rather than “pure, hard math and science,” she said.

What resources does AI need?

The quality of an AI model scales as its training data and computational resources (or “compute”) improve. “It’s an industry that requires a lot of capital to be successful,” Friar said.

In pre-training, for example, Friar said, a general-purpose model becomes “smarter” as it takes in more data across more compute. The subsequent stage might involve fine-tuning the model. “Let’s say I wanted to create models that are really good at diagnosing diseases,” she said. “So we ingest all the human information available out there for the general-purpose model. [But] for fine tuning, we’d say: ‘Now I want it to digest medical textbooks, I want it to get very deep in diagnostics.’”

“Compute” — the data centers and processors that crunch the data — are also a critical part of this training. In January, a consortium that includes OpenAI announced a plan to invest $500 billion by 2029 to build around 10 gigawatts (GW) worth of data centers — a power consumption load greater than the residential power consumption in Ireland. “We have to find sites, find power, dig holes in ground, put up data centers, fill them with equipment, buy GPUs [chips], and prove we can do this.”

What feels like an ambitious $500 billion investment today will soon be seen as just the beginning, Friar said. She added that we’re entering a generational opportunity to rebuild critical infrastructure — cleaner, smarter, and at global scale.

This moment calls for bold thinking: scaling renewables, reimagining the workforce, and redefining economic development zones, Friar said. It’s a chance to unlock new industries, create high-quality jobs, and strengthen global leadership. The future isn’t just about powering AI — it’s about powering progress.

 

This article is being provided for educational purposes only. The information contained in this article does not constitute a recommendation from any Goldman Sachs entity to the recipient, and Goldman Sachs is not providing any financial, economic, legal, investment, accounting, or tax advice through this article or to its recipient. Neither Goldman Sachs nor any of its affiliates makes any representation or warranty, express or implied, as to the accuracy or completeness of the statements or any information contained in this article and any liability therefore (including in respect of direct, indirect, or consequential loss or damage) is expressly disclaimed.